Get error when train TransformerXL on COOT embeddings

[menggehe]

I am trying to train TransformerXL model on video captioning task.
command:
python train_caption.py -c config/caption/paper2020/anet_coot_vidclip_mart.yaml --coot_feat_dir embeddings

But I get an error when validation:

Is there anything that you suggest I do?

[simon-ging]

Hello, we currently don't support running TransformerXL with this repository. Check the original MART code, there you can run it:
https://github.com/jayleicn/recurrent-transformer

If you get TransformerXL running with this repo, please feel free to provide a pull request.

[menggehe]

Thank you very much!